基于分子动力学及多智能体的地铁车站行人仿真研究

The pedestrian density of metro station is relatively high, and pedestrian often touch or even squeeze each other. Discrete micro-simulation model only study the movement rather than the force of the pedestrian, and also the attained pedestrian density in these models is smaller than that in practice, so these models are not very applicable for metro station. The social force model studies the pedestrian movement from the view of force, taking both the movement and the force into account. Therefore it can solve the above problem. However, there are still some problems for this model. This research will focus on these problems. Firstly, with the molecular dynamic method, the computation algorithm will be designed, and the linked list cell method will be used to accelerate the computing speed. Secondly, the intelligent route choice algorithm will be formulated based on multi-agent theory to eliminate the disadvantage that pedestrians are considered as physical particles. Then the pedestrian will be simulated as ellipse and the model parameters will be calibrated and verified through pedestrian field study to reach more realistic results. Finally, a pedestrian simulation tool will be developed and will be applied in layout evaluation of metro station facilities.

地铁车站行人密度较高，行人之间会发生接触甚至挤压现象，离散微观行人仿真方法只能研究行人的运动，不能反映其受力，仿真达到的最大行人密度与实际有差距，不能很好地用于车站行人仿真。社会力模型从受力的角度研究行人运动规律，兼顾了行人运动状态和受力状态，能解决上述问题，但是该模型还存行人路径选择盲目、发生不必要的碰撞等问题。本研究拟针对这些问题，采取分子动力学相关方法，设计社会力模型的分子动力学算法，引入linked-list cell方法，编程实现社会力模型，提高社会力模型计算速度；基于多智能体思想，设计行人智能路径选择算法，克服社会力模型将行人等同于物理粒子的弊端；采用椭圆形模拟行人，并基于车站行人调研和测试，标定校核模型参数，得到更符合实际的仿真效果；最终开发仿真工具，并将其应用于车站设施布局的仿真评价。

轨道交通站内行人与设施布局的互动性分析研究是目前轨道交通设计、运营管理的重要理论基础。社会力模型以模型简单、行人行为重现能力强、参数具有实际的物理意义、便于标定等优点成为主流模型之一。然而，该模型目前的研究中所采用的数值解法效率较低，不适合大规模行人的模拟研究；并且，模型将人作为物理粒子导致的运动盲目性，以及部分模型参数取值缺乏依据，导致模型满足国内站内高密度行人群模拟。因此，本研究采用分子动力学的多自由度微分方程的Gear预测校正算法对社会力模型进行求解，同时采用Linked-list Cell方法提高算法计算效率，并利用柯朗条件确定最优的计算时间步长。引入多智能体的思想，通过添加视域模块，使行人感知周围环境，并对环境信息做出行为反应；构建与行走环境互动的行人模型，更加真实地表现行人与站内设施环境的关系。另外，自主调研站内乘客的运动特征，标定模型中的运动参数；并且自主设计实验测定人体弹性模量，重新标定人体力学参数。基于上述工作借助C++编程语言初步开发地铁车站行人仿真工具，并车站设计方案及设施布局进行了初步的优化评价。本项目通过对社会力模型的改进，为轨道交通车站内高密度人群的仿真提供了系统的解决方法，对于站内设施的布局优化评价具有重要的意义。